FR3115913A1 - Method and system for automatically generating and using road traffic scenarios - Google Patents

Abstract

Method and system for automatically generating and using road traffic scenarios The invention relates to a method and a system for the automatic generation and use of road traffic scenarios, comprising one or more fixed sensors (4A, 4B, 4C), the or each sensor being adapted to acquire raw data relating to a perimeter of observation, the system (2) further comprising at least one calculation processor (10) configured to implement a module (16) for obtaining, from the raw data acquired by the sensor or sensors, a plurality of fixed or dynamic elements forming a succession of observed scenes, and parameters characterizing said elements, a module (18) for identifying events by comparison between at least two observed scenes of the succession of scenes, a module (20) for generating a concrete scenario, comprising a temporal sequence of scenes linked by events and a module (22) for updating a base of scenarios (14). Figure for abstract: Figure 1

Description

Method and system for automatically generating and using road traffic scenarios

The present invention relates to a method and a system for automatically generating and using road traffic scenarios.

The invention lies in the field of the secure piloting of motor vehicles, and in particular in the field of road safety. It finds an application in particular in the field of autonomous or semi-autonomous motor vehicles.

In the field of safe piloting of motor vehicles, and in particular in autonomous piloting, one of the main problems is the early identification of risky situations, making it possible to take corrective measures to avoid accidents.

The ISO/PAS 21448 standard defines the expected functional safety for road vehicles or SOTIF ( Safety Of The Intended Functionality ). For the implementation of autonomous or semi-autonomous vehicle functionalities, knowledge and anticipation of risky situations are necessary to ensure the prescribed level of safety. It is planned to use previously recorded scenarios, each scenario describing temporal sequences between scenes, observed to identify risk situations and take the necessary measures.

The document WO2017/164991 A1 describes a system that implements an autonomous vehicle that includes an on-board traffic scenario detector, in order to increase safety, in particular for the reliability of the detection of objects according to the traffic scenario. In this document, the traffic scenario is limited to the road configuration around the autonomous vehicle carrying the detector, such as the number of lanes, intersections, etc., without however making it possible to anticipate situations at risk of accident.

Thus, a problem that arises in the field of road safety for autonomous or semi-autonomous vehicles is to develop a database of road traffic scenarios that is as complete, as reliable as possible and which is also up-to-date. Indeed, the autonomous or semi-autonomous vehicles put into circulation for tests are few in number, and therefore the sensors on board these vehicles provide a relatively low hourly volume of observation, which makes it difficult to build a base of traffic scenarios, for a controlled infrastructure that is sufficiently reliable.

An alternative possibility is to generate road traffic scenarios from expert knowledge, but this solution is particularly laborious. In addition, manual generation of a database of road traffic scenarios does not allow high precision of the numerical values of the parameters defining each scenario, and therefore does not allow alerts to be raised with sufficient precision.

The object of the invention is to remedy the drawbacks of the state of the art, by allowing automatic generation of road traffic scenarios, which is precise and reliable, regularly updated and makes it possible to increase the level of traffic safety. road.

To this end, the invention proposes a method for the automatic generation and use of road traffic scenarios, implemented by a system comprising one or more fixed sensors, the or each sensor being suitable for acquiring raw data relating to a observation perimeter. This method comprises steps, implemented by a calculation processor, of:

• obtaining, from the raw data acquired by the sensor(s), a plurality of fixed or dynamic elements forming a succession of observed scenes, and parameters characterizing said elements,
• identification of events by comparison between at least two observed scenes of the succession of scenes,
• generation of a concrete scenario, comprising a temporal sequence of scenes linked by events;
• updating a base of scenarios using said generated concrete scenario.

Advantageously, the method of the invention makes it possible to generate and gradually enrich a database of road traffic scenarios, which then makes it possible to achieve a level of functional safety expected for autonomous or semi-autonomous vehicles .

The method for automatically generating and using road traffic scenarios according to the invention may have one or more of the characteristics below, taken independently or in all acceptable combinations.

Updating the scenario base includes comparing the concrete scenario generated with logical scenarios previously recorded in the scenario base, and, if said concrete scenario generated belongs to a logical scenario already stored, updating the statistics associated with said logical scenario.

Updating a scenario base includes comparing the concrete scenario generated with logical scenarios previously recorded in the scenario base, and, if said concrete scenario generated does not belong to any logical scenario already stored, modifying the base scenarios by adding a new corresponding logical scenario or by modifying parameters characterizing a previously stored logical scenario.

The method further comprises, prior to the addition of a new logical scenario in the base of scenarios, a step of validating said new logical scenario.

The method further comprises, following the addition of a new logical scenario in the scenario base, a verification of the scenario base to determine whether said new logical scenario belongs to any previously recorded functional scenario, and in the event of negative determination, a generation of a new functional scenario and an addition of said new functional scenario in the base of scenarios.

The method comprises, following the obtaining of raw data, a step of converting the raw data into parametrized data characterizing observed traffic elements.

The method also includes a fusion of parametrized data from several sensors and a calculation of parameters related to the observed elements.

The method further comprises a step of parameterizing an observed scene from the observed traffic elements and the calculated associated parameters, comprising the placement of the traffic elements on a map of the observed site containing the fixed elements of the observed site.

The method further includes a step of enriching the parametrization of the parameterized scene with data from external databases, to obtain an enriched scene.

The method further comprises detecting the criticality of at least one observed scene and, in the event of criticality greater than a threshold, sending an alert to autonomous or semi-autonomous vehicles of the observed site.

According to another aspect, the invention relates to a computer program comprising software instructions which, when executed by a processor of a programmable device, implement a method of automatically generating and using scenarios road traffic as briefly described above.

According to another aspect, the invention relates to a non-volatile information recording medium on which is recorded a computer program comprising software instructions which, when they are executed by a processor of a programmable device, put implements a method of automatically generating and using road traffic scenarios as briefly described above.

According to another aspect, the invention relates to a system for the automatic generation and use of road traffic scenarios, comprising one or more fixed sensors, the or each sensor being suitable for acquiring raw data relating to an observation perimeter, the system further comprising at least one calculation processor configured to implement:

• a module for obtaining, from the raw data acquired by the sensor(s), a plurality of fixed or dynamic elements forming a succession of observed scenes, and parameters characterizing said elements,
• an event identification module by comparison between at least two observed scenes of the succession of scenes,
• a module for generating a concrete scenario, comprising a temporal sequence of scenes linked by events;
• a module for updating a base of scenarios using said concrete scenario used.

The system for automatically generating and using road traffic scenarios is configured to implement a method for automatically generating and using road traffic scenarios, according to all the variants briefly explained above.

Other characteristics and advantages of the invention will emerge from the description given below, by way of indication and in no way limiting, with reference to the appended figures, among which:

there schematically illustrates a system for automatically generating and using road traffic scenarios according to one embodiment of the invention;

there schematically illustrates a succession of road traffic scenes forming a road traffic scenario;

there is a block diagram of the main steps of a method for automatically generating and using road traffic scenarios according to one embodiment of the invention.

There schematically illustrates a system 2 for automatically generating and using road traffic scenarios, relative to an observed site. This system comprises sensors 4A, 4B, 4C which are fixed and are positioned on the observed site (not shown in ). Each sensor is configured to acquire raw data relating to an observation perimeter belonging to the site.

The sensors 4A, 4B, 4C are for example optical cameras, or LIDAR-type scanning lasers, or any other type of detection radar, or even a mixture of these various types of sensors.

The sensors 4A, 4B, 4C provide raw observation data, with a given temporal frequency.

The raw data is transmitted, by an appropriate means of communication, to a system 6 for monitoring and calculation.

For example, the raw observation data are communicated by means of a wireless communication, for example according to the WiFi communication protocol or according to the Bluetooth communication protocol.

According to a variant, the raw observation data are communicated via a wired link, for example an Ethernet link.

The monitoring and calculation system 6 comprises a communication interface 8, suitable for communicating with the sensors 4A, 4B, 4C, in particular suitable for receiving the raw observation data, depending on the means of communication, wireless or wired, chosen. .

In one embodiment, the monitoring and calculation system 6 forms part of a centralized control and command post (PCC), which is configured to communicate with a plurality of autonomous or semi-autonomous motor vehicles, according to a protocol predetermined radio communication, according to a two-way communication mode. For example, the 4G communication standard is used. In particular, the PCC is suitable for transmitting alerts to autonomous or semi-autonomous vehicles, in the event of detection of a risky situation, in order to prevent an accident, as detailed below. In particular, thanks to the invention and as explained in more detail below, the transmission of alerts is carried out by automatic processing using the generated and recorded traffic scenarios.

According to a variant, the communication is carried out using a traveler information and operation assistance system (SAEIV).

The monitoring and calculation system 6 further comprises a calculation unit 10, for example one or more calculation processor(s) and an electronic memory unit 12, adapted to memorize program code instructions executable by the processor 10 For example, the electronic memory 12 is a ROM memory or a RAM memory.

According to one embodiment, the calculations are performed on one or more calculation processors 10, for example in a distributed system of the “cloud computing” type.

As a variant, the electronic memory 12 is an optical disc, a magneto-optical disc, any type of non-volatile memory (for example EPROM, EEPROM, FLASH, NVRAM), a magnetic card or an optical card, or any medium capable of store electronic instructions and to be coupled to a computer system bus.

The electronic memory forms a computer-readable medium, for example a medium, suitable for memorizing program code instructions making it possible to implement a method for automatically generating and using road traffic scenarios according to the invention.

In addition, the system 6 comprises or is connected to a scenario base 14, which stores traffic scenarios, formed from elements observed, as explained in more detail below.

The elements, in particular traffic elements of the observed site, are for example safety guardrails, sidewalks, traffic signs, pedestrians, parked or moving vehicles, for example bicycles, motor vehicles, buses, etc.

A scene is made up of a set of traffic elements observed at an instant in time. The elements include, for example, fixed elements of the observed environment (roads, with intersections, roundabouts, information panels, light signals such as traffic lights, and other fixed elements), as well as dynamic elements, including the presence is temporary, eg road users – motor vehicles, bicycles, pedestrians, others (eg animals, various obstacles, eg boxes, tires etc).

The observed elements have an associated state.

For example, for dynamic elements, the associated state includes the position of the dynamic element, the direction of movement, the speed of movement.

Preferably, the scenario base 14 is a hierarchical database, the traffic scenarios being organized into categories hierarchically at three levels of hierarchy.

The first level of hierarchy includes functional scenarios.

The second level of hierarchy includes logical scenarios.

The third level of hierarchy includes concrete scenarios.

A traffic scenario is defined from a succession of traffic scenes observed in a chosen time interval.

By concrete scenario, we mean a real, that is to say observed, traffic scenario, in which each scene is characterized by recorded data, for example the values of the states associated with the dynamic elements of the scene. Moreover, a concrete scenario comprises events calculated from the comparison between at least two observed scenes of the succession of scenes. For example, an event is defined by an action of moving a dynamic element with its characteristics.

A logical scenario is understood to mean a traffic scenario defined by events associated with ranges of values, certain precise characteristics of the concrete scenarios being omitted or grouped into more general categories. In other words, a logical scenario is a generalization of concrete scenarios. Thus, a logical scenario is a class of concrete scenarios.

By functional scenario, we mean a scenario at the highest level of abstraction in which the events are described by one or more sentences. Thus, a functional scenario is a class of logical scenarios.

For example, the illustrates the same site, which is a road roundabout 30 with four entrances/exits 321, 322, 323, 324, each entrance/exit being connected to a corresponding road with two lanes, corresponding to two opposite directions of traffic. Entry and exit are defined by the expected direction of traffic.

There illustrates three scenes 34, 36, 38, each corresponding to an observation of the site at a given time instant in a succession of time instants.

The scene 34 comprises the entry of a motor vehicle 40 on the entry 32 ₁ of the roundabout 30, in the intended direction of traffic. The scene 36 comprises the passage of the motor vehicle 40 on the roundabout 30, opposite the entrance/exit 32 ₂ . The scene 38 comprises the exit of the motor vehicle 40 from the roundabout 30 onto the exit lane from the entrance/exit 32 ₃ .

A concrete scenario corresponding to the observation of scenes 34, 36, 38 is for example defined by: passage of a city type car from entry 32 ₁ to exit 32 ₃ , with braking at 12.3 km/h in and out at 25.4 km/h.

A corresponding logical scenario is for example the passage of a car, entry 32 ₁ , exit 32 ₃ , entry speed comprised in the range of 5 to 30 km/h and exit speed comprised in the range of 5 to 30 km/h. h.

A corresponding functional scenario is for example defined by "passage of a car on the roundabout at nominal speed, between 5 and 30 km/h".

The traffic scenario given as an example here is a nominal traffic scenario, the vehicle 40 traveling at an adequate speed, in the direction of the traffic, which does not present any criticality or risk of associated accident.

Of course, the traffic scenario of the has been given by way of example, and it is understood that a large number of scenarios are possible, involving one or more other vehicles, for example cars, trucks, buses, with other roundabout routes, crossings, bicycle or pedestrian crossings…

Back to the , in one embodiment, the electronic memory 12 stores:

a module 16 for obtaining, from the raw data acquired by the sensor or sensors, a plurality of fixed or dynamic elements forming a succession of observed scenes, and parameters characterizing said elements;

a module 18 for analyzing the observed scenes to identify events between at least two observed scenes of the succession of scenes,

- an associated logical scenario generation module 20,

- a module 22 for updating the base of scenarios by using the logical scenario generated;

- a module 24 for critical scene detection and alert escalation in the event of critical scene detection.

In one embodiment, the modules 16, 18, 20, 22 and 24 are implemented in the form of software and form a computer program which implements a method of automatically generating and using road traffic scenarios according to a embodiment of the invention when it is executed by a processor of a programmable device, for example a computer or a set of computers of a computer system.

In a variant not shown, the modules 16, 18, 20, 22 and 24 are each made in the form of a programmable logic component, such as an FPGA ( Field Programmable Gate Array ), a GPU (graphics processor) or a GPGPU ( General-purpose processing on graphics processing ), or even in the form of a dedicated integrated circuit, such as an ASIC ( Application Specific Integrated Circuit ).

There is a block diagram of the main steps of a method for automatically generating and using road traffic scenarios, according to one embodiment of the invention.

The method includes a first step 50 of receiving the raw data captured by the various sensors which observe the site, relating to the observed traffic.

For example, raw data is heterogeneous data including video obtained by cameras or point clouds obtained by lidar.

These raw data are processed, the processing comprising a conversion 52 into parametrized data. The parameterized data includes values of speed, dimension (i.e. height and/or width and/or depth) according to categories of traffic elements.

The parameterized data is calculated and stored in an appropriate format to be then processed automatically. In other words, parameterized data is data that can be interpreted by a machine, but which is not usable or interpretable by a human operator.

In one embodiment, using algorithms, observed elements are extracted from the raw data, in particular traffic elements: motor vehicles, pedestrians, bicycles. For example, a convolutional neural network is implemented for the recognition and classification of observed traffic elements. In addition, the dimensions of the detected traffic elements are calculated, for example using modules of OpenCV (library of real-time image processing operations).

According to a variant, steps 50 and 52 are carried out at the same time, using specific sensors for recognizing road traffic elements.

At the end of step 52, a list of traffic elements present in each scene observed is obtained, also called a monitoring list.

The monitoring lists associated with successive scenes make it possible to monitor the elements present, for example the movement of a car or a pedestrian.

In one embodiment, a tracking list comprises descriptive information of the elements present in the scene, in the form of parametrized data, comprising, for example: classification, dimensions, position, speed of movement, direction of movement.

A monitoring list contains parametrized data, which is formatted and automatically exploitable by machine processing.

When several elements of the same category are present, they are preferably identified by an identifier in the tracking list, for example a number.

Step 52 is followed by a step 54 for deleting irrelevant data, for example in the case where the observed scene is frozen. Freeze scene detection is performed by automatic processing of recorded data. More generally, the determination of irrelevant data is carried out by automatic processing on the parameterized data recorded.

In this case the monitoring list is not kept, and step 54 is followed by step 50 previously described.

If there remain relevant data, i.e. moving traffic elements, in the tracking list, step 54 is optionally followed by a merging step 56, consisting in merging the elements observed by nature sensors. different or in a different position to obtain an enriched scene. For example, certain traffic elements are observed by certain sensors only. A succession of enriched scenes includes all the traffic elements observed in a given time interval.

In step 56 a Kalmann filtering on the videos coming from several cameras is for example implemented.

Parameters linked to the traffic elements detected are then calculated in the parameter calculation step 58. The parameters are linked to the traffic elements observed and confirmed by previous observations, for example by comparing their values.

Thus, consolidated parameters are obtained.

Optionally, additional parameters are added in the tracking list, for example the acceleration of a moving traffic element.

The parameter calculation step 58 is followed by a parameterization step 60 of the scene observed from the identified traffic elements and the associated parameters, by placing these traffic elements on a map of the observed site containing the fixed elements of the site observed. For example, the observed site is the roundabout illustrated in , and the identified traffic element is motor vehicle 40.

The method also includes a step 62, which is optional, of enriching the parameterization of the scene with data from external databases, for example meteorological data, road traffic data near the observed site, information of the calendar (day of the week, calendar period, for example school holidays). The external data is for example received from data servers 64 using radio communication.

Finally, the enriched scene is analyzed during an analysis step 66 to determine its criticality. The criticality analysis of the scene, carried out automatically, uses the base of scenarios 14.

The scenario base 14 contains criticality levels associated with scenes with respect to scenarios already identified as being scenarios at risk of a potential accident. For example, this is the case if a vehicle is traveling in the opposite direction of traffic, when the speed of a vehicle is greater than an upper limit of a nominal speed range or when the traffic density exceeds a threshold, for example in connection with observed meteorological conditions (rain, fog, etc.).

A critical scene is a scene which includes at least one element having an associated parameter or a combination of parameters determined as critical, for example a car which has a speed of 70km/h on a road limited to 50km/h or if a car is driving against the direction.

For example, a maximum level of criticality is indicated if a car is driving in the wrong direction on a traffic lane.

For example, for a speed overrun, the level of criticality is based on a percentage of overrun of the maximum speed authorized or recorded in the corresponding traffic scenario. In one embodiment, the predetermined threshold is a percentage P of exceeding this maximum speed, P being for example equal to 5%.

In the event of detected criticality, that is to say if the level of criticality is greater than a predetermined threshold, an alert raised during an alert raising step 68.

For example, an alert notification is automatically transmitted to a guidance module of an autonomous vehicle.

Thus, advantageously, the criticality is detected automatically using the scenario database, and the alert is raised automatically, which makes it possible to accelerate the raising of the alert and consequently to increase road safety.

Step 62 is also followed by a step 70 of identifying events characterizing a succession of enriched scenes, by comparing the values of the parameters or of the states characterizing the scenes. Event detection is used to detect events such as the entry of a vehicle at a roundabout entrance, and the exit of the vehicle at a roundabout exit.

Step 70 of identifying events is followed by a step 72 of generating a concrete scenario or several concrete scenarios from the succession of scenes and associated events.

In one embodiment, a concrete scenario is calculated for the duration of observation of an element in the observed perimeter. If a given scene contains several moving elements, a concrete scenario is generated for each moving element of the scene.

Subsequently, it is considered that a concrete scenario is obtained in step 72, but the method applies analogously when several concrete scenarios are generated.

The concrete scenario obtained in step 72 is defined by elements and events, and associated parameter values, for example the type of element (e.g. pedestrian, bicycle, motor vehicle and type of motor vehicle), the speed of movement, the way of movement etc.

It is then sought, in the classification step 74, to classify the concrete scenario with respect to the previously recorded logical scenarios. In particular, it is determined whether the elements and events defining the concrete scenario belong to a registered logical scenario.

It is checked in the test step 76 whether the concrete scenario obtained belongs to a logical scenario present in the base of scenarios, by comparing the concrete scenario with the logical scenarios recorded in the base.

For example, if the concrete scenario indicates a motor vehicle which enters the roundabout by a first entry/exit and exits by a second entry/exit, with a given speed of crossing the roundabout, it is checked whether there is a logical scenario already saved with compatible parameters.

For example, if the logical scenario with traffic between the first entry/exit and the second entry/exit is recorded, with a roundabout crossing speed between 15km/h and 20km/h, and the concrete scenario indicates a speed of crossing the roundabout of 20 km/h, then the concrete scenario observed belongs to this logical scenario already recorded in the database, therefore the answer to the test of step 76 is “yes”. On the other hand, if the speed at which the car crosses the roundabout is 27 km/h, this speed is outside the previously recorded range, so the answer to test 76 is “no”.

If the answer to step 76 is "yes", step 76 is followed by step 84 of updating the statistics of the logical scenario in the scenario database.

On the other hand, if the answer to step 76 is “no”, this step is optionally followed by a step 78 of validation of the new logical scenario by an expert.

According to a variant, automatic algorithmic processing is implemented to perform this validation.

In particular, this step makes it possible to determine whether the concrete scenario determined in step 72 belongs to a new class of scenarios, i.e. a new logical scenario which was not previously recorded in the scenario base, and whether the new logical scenario belongs to an existing class of logical scenarios (i.e. functional scenario), or whether a new corresponding functional scenario should be defined and recorded. In addition, a level of criticality associated with one or more scenes of the scenario, for example with respect to an accident risk, is calculated and stored at this stage. For example, when the observed speed of movement of the vehicle exceeds the maximum limit of the nominal speed range, the percentage of excess is used to calculate a level of criticality, the higher this percentage, the higher the risk of accident. . For example, the level of criticality is defined by an integer value coded on a given scale, for example between 1 and N, N being an integer and corresponding to the maximum level of criticality.

In the event of a positive determination (step 80), a new logical scenario, and, where applicable, a new functional scenario, is added to the scenario base during a recording step 82.

In the event of a negative determination, step 80 is followed by step 84 of updating the statistics of the logical scenario. Continuing the previous example, if the speed of crossing the roundabout is 27 km/h, but this speed is acceptable, it is possible to modify the corresponding logical scenario, by modifying the speed interval which becomes 15 to 27 kph. If necessary, the level of criticality is also re-calculated and stored for one or more scenes of the scenario.

Steps 74 to 84 carry out an update of the scenario base by using the logical scenario determined from the concrete scenario.

Advantageously, the invention allows the automatic and regular enrichment of the base of scenarios, by regular observation of road conditions. Thus, the scenario database is constantly updated automatically.

Advantageously, the invention allows the automatic generation and use of road traffic scenarios, in particular for increasing road safety, when operating autonomous or semi-autonomous vehicles.

The method of the invention applies analogously to different sites, so as to form a base of multi-site scenarios.

Claims (13)

Method for automatically generating and using road traffic scenarios, implemented by a system (2) comprising one or more fixed sensors (4A, 4B, 4C), the or each sensor (4A, 4B, 4C) being adapted to acquire raw data relating to an observation perimeter, the method being characterized in that it comprises steps, implemented by a calculation processor, of:

• obtaining (50-60), from the raw data acquired by the sensor or sensors, a plurality of fixed or dynamic elements forming a succession of observed scenes, and parameters characterizing said elements,
• identification (70) of events by comparison between at least two observed scenes of the succession of scenes,
• generation (72) of a concrete scenario, comprising a temporal sequence of scenes linked by events;
• updating (74-84) of a base of scenarios (14) using said generated concrete scenario.

Method according to claim 1, in which the updating (74-84) of the scenario base (14) comprises a comparison (74, 76) of the concrete scenario generated with logical scenarios previously recorded in the scenario base (14 ), and, if said concrete scenario generated belongs to a logical scenario already stored, updating (84) the statistics associated with said logical scenario. Method according to claim 1, in which the updating (74-84) of a scenario base (14) comprises a comparison of the concrete scenario (74, 76) generated with logical scenarios previously recorded in the scenario base ( 14), and, if said concrete scenario generated does not belong to any logical scenario already stored, modification (78-82) of the base of scenarios by adding a new corresponding logical scenario or by modifying parameters characterizing a logical scenario previously memorized. Method according to claim 3, further comprising, prior to the addition of a new logical scenario in the base of scenarios, a validation step (78) of said new logical scenario. Method according to claim 3 or 4, further comprising, following the addition of a new logical scenario in the scenario base, a verification of the scenario base to determine whether the said new logical scenario belongs to a previously recorded functional scenario , and in the event of a negative determination, a generation of a new functional scenario and an addition of said new functional scenario in the scenario base (14). Method according to any one of Claims 1 to 5, comprising, following the obtaining (50) of raw data, a step of converting (52) the raw data into parametrized data characterizing observed traffic elements. Method according to claim 6, further comprising a fusion (56) of the parametrized data coming from several sensors and a calculation of parameters linked to the observed elements. Method according to one of Claims 1 to 7, further comprising a step of parametrizing (60) an observed scene from the observed traffic elements and the calculated associated parameters, comprising placing the traffic elements on a map of the observed site containing the fixed elements of the observed site. Method according to claim 8, further comprising a step of enriching (62) the parameterization of the parameterized scene with data from external databases, to obtain an enriched scene. Method according to one of Claims 1 to 9, comprising a detection (66) of criticality of at least one observed scene and, in the event of criticality greater than a threshold, an automatic sending (68) of an alert to autonomous vehicles or semi-autonomous from the observed site. Computer program comprising software instructions which, when executed by a processor of a programmable device, implement a method of automatically generating and using road traffic scenarios according to claims 1 to 10. System for the automatic generation and use of road traffic scenarios, comprising one or more fixed sensors (4A, 4B, 4C), the or each sensor being suitable for acquiring raw data relating to an observation perimeter, the system ( 2) being characterized in that it further comprises at least one calculation processor (10) configured to implement:

• a module (16) for obtaining, from the raw data acquired by the sensor or sensors, a plurality of fixed or dynamic elements forming a succession of observed scenes, and parameters characterizing said elements,
• a module (18) for identifying events by comparison between at least two observed scenes of the succession of scenes,
• a module (20) for generating a concrete scenario, comprising a temporal sequence of scenes linked by events;
• a module (22) for updating a base of scenarios.

System for the automatic generation and use of road traffic scenarios according to claim 12, further comprising a module (24) for detecting a critical scene and raising the alert to autonomous or semi-autonomous vehicles of the site observed in the event of critical scene detection.